Understanding Brain Functional Dynamics Through Neural Koopman Operator With Control Mechanism.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Understanding how cognition and behavior arise from brain function is a fundamental neuroscience challenge.
• Existing computational models struggle to capture the complex, non-linear dynamics of the human brain, often resorting to simplified linear models or opaque neural networks.
• The human brain is a complex biological system with non-linear and self-organized dynamics.

Purpose of the Study:

• To develop an end-to-end deep model for identifying underlying brain dynamics.
• To leverage Koopman operator theory for modeling complex non-linear systems in a linear space.
• To predict cognitive states from neuroimaging data by identifying latent dynamic systems.

Main Methods:

• Devised an end-to-end deep model based on Koopman operator theory.
• Incorporated a biology-inspired control module for feedback-based input adjustment.
• Applied the model to large-scale neuroimaging data to identify functional fluctuation dynamics.

Main Results:

• Successfully predicted cognitive states from neuroimaging data.
• Identified latent dynamic systems governing functional fluctuations in the brain.
• Demonstrated the model's potential for system-level understanding of brain-cognition relationships.

Conclusions:

• The developed deep model effectively identifies brain dynamics using Koopman operator theory.
• This approach facilitates a system-level understanding of the intricate relationship between brain function and cognition.
• The explainable deep models offer promising avenues for future neuroscience research.